Abstract
As traditional pollinators face increasing threats from climate change, the development of robotic pollination technology has become imperative, with apple flower detection emerging as a critical component of the technology. Deep learning (DL) advancements present novel methods in enhancing apple flower detection efficiency. However, deploying in real time on resource-constrained drone platforms demands a balance between computational efficiency and accuracy. To address this challenge, this study introduces an improved you-only-look-once version 5 small (YOLOv5s-Im) model by improving the original YOLOv5s architecture, using MobileNet version 3 as the backbone and GhostNet as the neck. This study then validated the YOLOv5s-Im performance by deploying it in real time on a drone platform designed for apple flower pollination. YOLOv5s-Im achieved an 88% detection accuracy and averaged 41.6 pollination attempts per 3-minute flight across five tests, significantly outperforming YOLOv5s and YOLOv5s with Transformers (YOLOv5s-T) as backbone (fewer than 10 attempts), due to its 2 FPS inference speed versus their 0.05 FPS. Control tests with lightweight models YOLOv5s with ShuffleNet version 2 (YOLOv5-Sh-V2) and YOLOv5s with MobileNet version 2 (YOLOv5s-Mb-V2) as backbones, averaged 37.8 and 30.6 attempts per flight, respectively, with accuracies of 80% and 82% mAP and detection speeds of 1.0 FPS and 0.7 FPS, further confirming YOLOv5s-Im’s superior balance of accuracy and efficiency. Its robust accuracy (84%-88%) across diverse conditions—clear light (88%), afternoon settings (86%), angled views (87%), and low-light shadows (84%)—demonstrates reliability in varied orchard environments. Compared to YOLOv5s, YOLOv5s-T, YOLOv7, YOLOv8, and Faster-R-CNN, YOLOv5s-Im excels with precision (90.6%), recall (87.7%), mAP50 (91.2%), and F1-score (89.42%), while reducing GFLOPS by 89% and model size by 85%, achieving high frame rates (227 FPS on NVIDIA RTX 4060 Ti, 22 FPS on Jetson Xavier, 4.56 FPS on Intel NUC11TNKi3). These results make YOLOv5s-Im an effective solution for real-time apple flower detection under natural lighting conditions in drone-based pollination systems.
Abstract from the paper’s OpenAlex record; open copy licensed cc-by. Read at the source ↗
The text the model read, with the source record’s inline markup (italics, sub- and superscripts) rendered and its character entities decoded; quote offsets refer to the raw text.
Claims 19
Claim · model-written summary
YOLOv5s-Im achieved 88% apple flower detection accuracy.
- Taxon
- apple
Malus domesticat0 - Outcome
- detection accuracye1
- Value
- 88%e1
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e1“YOLOv5s-Im achieved an 88% detection accuracy”
t0“apple flower detection”
Claim · model-written summary
YOLOv5s-Im averaged 41.6 pollination attempts per flight, outperforming YOLOv5s and YOLOv5s-T (fewer than 10 attempts).
- Taxon
- appleMalus domesticat0
- Outcome
- pollination attemptse2
- Value
- 41.6e2
- Comparator
- YOLOv5s and YOLOv5s with Transformers (YOLOv5s-T) as backbonee2
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e2“averaged 41.6 pollination attempts per 3-minute flight across five tests, significantly outperforming YOLOv5s and YOLOv5s with Transformers (YOLOv5s-T) as backbone (fewer than 10 attempts), due to its 2 FPS inference speed versus their 0.05 FPS”
t0“apple flower detection”
Claim · model-written summary
The advantage of YOLOv5s-Im in pollination attempts is attributed to its faster inference speed.
- Taxon
- appleMalus domesticat0
- Outcome
- pollination attemptse2
- Qualifier
- 2 FPS inference speed versus their 0.05 FPSe2
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e2“averaged 41.6 pollination attempts per 3-minute flight across five tests, significantly outperforming YOLOv5s and YOLOv5s with Transformers (YOLOv5s-T) as backbone (fewer than 10 attempts), due to its 2 FPS inference speed versus their 0.05 FPS”
t0“apple flower detection”
Claim · model-written summary
YOLOv5-Sh-V2 averaged 37.8 pollination attempts per flight with ShuffleNet version 2 backbone.
- Taxon
- appleMalus domesticat0
- Treatment
- ShuffleNet version 2e3
- Outcome
- attempts per flighte3
- Value
- 37.8e3
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e3“Control tests with lightweight models YOLOv5s with ShuffleNet version 2 (YOLOv5-Sh-V2) and YOLOv5s with MobileNet version 2 (YOLOv5s-Mb-V2) as backbones, averaged 37.8 and 30.6 attempts per flight, respectively, with accuracies of 80% and 82% mAP and detection speeds of 1.0 FPS and 0.7 FPS”
t0“apple flower detection”
Claim · model-written summary
YOLOv5-Sh-V2 reached 80% detection accuracy (mAP).
- Taxon
- appleMalus domesticat0
- Treatment
- ShuffleNet version 2e3
- Outcome
- accuraciese3
- Value
- 80%e3
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e3“Control tests with lightweight models YOLOv5s with ShuffleNet version 2 (YOLOv5-Sh-V2) and YOLOv5s with MobileNet version 2 (YOLOv5s-Mb-V2) as backbones, averaged 37.8 and 30.6 attempts per flight, respectively, with accuracies of 80% and 82% mAP and detection speeds of 1.0 FPS and 0.7 FPS”
t0“apple flower detection”
Claim · model-written summary
YOLOv5-Sh-V2 ran at 1.0 FPS detection speed.
- Taxon
- appleMalus domesticat0
- Treatment
- ShuffleNet version 2e3
- Outcome
- detection speedse3
- Value
- 1.0 FPSe3
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e3“Control tests with lightweight models YOLOv5s with ShuffleNet version 2 (YOLOv5-Sh-V2) and YOLOv5s with MobileNet version 2 (YOLOv5s-Mb-V2) as backbones, averaged 37.8 and 30.6 attempts per flight, respectively, with accuracies of 80% and 82% mAP and detection speeds of 1.0 FPS and 0.7 FPS”
t0“apple flower detection”
Claim · model-written summary
YOLOv5s-Mb-V2 averaged 30.6 pollination attempts per flight with MobileNet version 2 backbone.
- Taxon
- appleMalus domesticat0
- Treatment
- MobileNet version 2e3
- Outcome
- attempts per flighte3
- Value
- 30.6e3
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e3“Control tests with lightweight models YOLOv5s with ShuffleNet version 2 (YOLOv5-Sh-V2) and YOLOv5s with MobileNet version 2 (YOLOv5s-Mb-V2) as backbones, averaged 37.8 and 30.6 attempts per flight, respectively, with accuracies of 80% and 82% mAP and detection speeds of 1.0 FPS and 0.7 FPS”
t0“apple flower detection”
Claim · model-written summary
YOLOv5s-Mb-V2 reached 82% detection accuracy (mAP).
- Taxon
- appleMalus domesticat0
- Treatment
- MobileNet version 2e3
- Outcome
- accuraciese3
- Value
- 82%e3
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e3“Control tests with lightweight models YOLOv5s with ShuffleNet version 2 (YOLOv5-Sh-V2) and YOLOv5s with MobileNet version 2 (YOLOv5s-Mb-V2) as backbones, averaged 37.8 and 30.6 attempts per flight, respectively, with accuracies of 80% and 82% mAP and detection speeds of 1.0 FPS and 0.7 FPS”
t0“apple flower detection”
Claim · model-written summary
YOLOv5s-Mb-V2 ran at 0.7 FPS detection speed.
- Taxon
- appleMalus domesticat0
- Treatment
- MobileNet version 2e3
- Outcome
- detection speedse3
- Value
- 0.7 FPSe3
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e3“Control tests with lightweight models YOLOv5s with ShuffleNet version 2 (YOLOv5-Sh-V2) and YOLOv5s with MobileNet version 2 (YOLOv5s-Mb-V2) as backbones, averaged 37.8 and 30.6 attempts per flight, respectively, with accuracies of 80% and 82% mAP and detection speeds of 1.0 FPS and 0.7 FPS”
t0“apple flower detection”
Claim · model-written summary
YOLOv5s-Im reached 90.6% precision.
- Taxon
- appleMalus domesticat0
- Outcome
- precisione4
- Value
- 90.6%e4
- Comparator
- YOLOv5s, YOLOv5s-T, YOLOv7, YOLOv8, and Faster-R-CNNe4
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e4“Compared to YOLOv5s, YOLOv5s-T, YOLOv7, YOLOv8, and Faster-R-CNN, YOLOv5s-Im excels with precision (90.6%), recall (87.7%), mAP50 (91.2%), and F1-score (89.42%)”
t0“apple flower detection”
Claim · model-written summary
YOLOv5s-Im reached 87.7% recall.
- Taxon
- appleMalus domesticat0
- Outcome
- recalle4
- Value
- 87.7%e4
- Comparator
- YOLOv5s, YOLOv5s-T, YOLOv7, YOLOv8, and Faster-R-CNNe4
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e4“Compared to YOLOv5s, YOLOv5s-T, YOLOv7, YOLOv8, and Faster-R-CNN, YOLOv5s-Im excels with precision (90.6%), recall (87.7%), mAP50 (91.2%), and F1-score (89.42%)”
t0“apple flower detection”
Claim · model-written summary
YOLOv5s-Im reached 91.2% mAP50.
- Taxon
- appleMalus domesticat0
- Outcome
- mAP50e4
- Value
- 91.2%e4
- Comparator
- YOLOv5s, YOLOv5s-T, YOLOv7, YOLOv8, and Faster-R-CNNe4
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e4“Compared to YOLOv5s, YOLOv5s-T, YOLOv7, YOLOv8, and Faster-R-CNN, YOLOv5s-Im excels with precision (90.6%), recall (87.7%), mAP50 (91.2%), and F1-score (89.42%)”
t0“apple flower detection”
Claim · model-written summary
YOLOv5s-Im reached an F1-score of 89.42%.
- Taxon
- appleMalus domesticat0
- Outcome
- F1-scoree4
- Value
- 89.42%e4
- Comparator
- YOLOv5s, YOLOv5s-T, YOLOv7, YOLOv8, and Faster-R-CNNe4
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e4“Compared to YOLOv5s, YOLOv5s-T, YOLOv7, YOLOv8, and Faster-R-CNN, YOLOv5s-Im excels with precision (90.6%), recall (87.7%), mAP50 (91.2%), and F1-score (89.42%)”
t0“apple flower detection”
Claim · model-written summary
YOLOv5s-Im reduced GFLOPS by 89%.
- Taxon
- appleMalus domesticat0
- Outcome
- GFLOPSe5
- Value
- 89%e5
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e5“while reducing GFLOPS by 89% and model size by 85%”
t0“apple flower detection”
Claim · model-written summary
YOLOv5s-Im reduced model size by 85%.
- Taxon
- appleMalus domesticat0
- Outcome
- model sizee5
- Value
- 85%e5
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e5“while reducing GFLOPS by 89% and model size by 85%”
t0“apple flower detection”
Claim · model-written summary
YOLOv5s-Im ran at 227 FPS on an NVIDIA RTX 4060 Ti.
- Taxon
- appleMalus domesticat0
- Outcome
- high frame ratese6
- Value
- 227 FPSe6
- Qualifier
- on NVIDIA RTX 4060 Tie6
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e6“achieving high frame rates (227 FPS on NVIDIA RTX 4060 Ti, 22 FPS on Jetson Xavier, 4.56 FPS on Intel NUC11TNKi3)”
t0“apple flower detection”
Claim · model-written summary
YOLOv5s-Im ran at 22 FPS on a Jetson Xavier.
- Taxon
- appleMalus domesticat0
- Outcome
- high frame ratese6
- Value
- 22 FPSe6
- Qualifier
- on Jetson Xaviere6
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e6“achieving high frame rates (227 FPS on NVIDIA RTX 4060 Ti, 22 FPS on Jetson Xavier, 4.56 FPS on Intel NUC11TNKi3)”
t0“apple flower detection”
Claim · model-written summary
YOLOv5s-Im ran at 4.56 FPS on an Intel NUC11TNKi3.
- Taxon
- appleMalus domesticat0
- Outcome
- high frame ratese6
- Value
- 4.56 FPSe6
- Qualifier
- on Intel NUC11TNKi3e6
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e6“achieving high frame rates (227 FPS on NVIDIA RTX 4060 Ti, 22 FPS on Jetson Xavier, 4.56 FPS on Intel NUC11TNKi3)”
t0“apple flower detection”
Claim · model-written summary
YOLOv5s-Im kept 84%-88% accuracy across clear light, afternoon, angled and low-light orchard conditions.
- Taxon
- appleMalus domesticat0
- Outcome
- robust accuracye7
- Value
- 84%-88%e7
- Qualifier
- across diverse conditionse7
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e7“Its robust accuracy (84%-88%) across diverse conditions—clear light (88%), afternoon settings (86%), angled views (87%), and low-light shadows (84%)”
t0“apple flower detection”